July 11, 2020
Role of demand planning & forecasting in e-commerce
E-commerce has seen a spike in the recent past and continues to grow rapidly. The Pandemic situation has opened up new opportunities and companies are either scrambling to get on board or optimize their operations. According to Statista e-retail revenues are projected to grow to 6.54 trillion US dollars by 2022.
E-commerce sites, quintessentially, thrive on 3 key factors. Availability, Shipping or order fulfillment, and customer satisfaction. Having adequate stocks and being able to ship them out fast either make or break the sales for most online shops. According to research, 69% of shoppers abandon repeat purchase if goods were not delivered within 2 days of the promised date. Worst yet, 38% of shoppers will not purchase at all if the order takes longer than a week.
It is clear that inventory management is key to a successful e-retail. The challenge for optimal stock becomes far greater if your e-business sold highly perishable goods such as food items or medication.
Proper inventory management, which leads to availability is intertwined with consumer demand and requires forecasting at granular frequencies as opposed to quarterly or monthly levels.
What drives the demand in e-commerce?
The increasing service levels such as same-day delivery, a replacement for returns, pairing complementary products are influencing retailers to invest time and effort in inventory management. According to an analysis done by Beroe Inc, a procurement intelligence firm the market for warehousing is growing at an annual rate of 6-8%.
The e-commerce market is a highly competitive one. Whilst convenience is the fundamental demand driver in e-commerce, consumers dictate regular price offs, bundled offers, free delivery, and other means of incentive to shop at a particular e-shop. In the case of perishables freshness, fast delivery and substitute offers can influence the buy and repeat purchase.
Why large stocks and wastage?
Availability plays a key role in e-commerce. E-retailers are well aware of the costs of going Out of Stock; dissatisfied customers which translate into a loss of sales and market share to competitors.
Keeping buffer stocks to tackle this issue is a common practice amongst many but this leads to other related problems that may not seem as critical as losing a sale, at first.
Since E- retailers are required to maintain a healthy stock of a wide variety of items that may be sold 24x 7x 365, warehousing is crucial. In addition, proper storage to minimize damage and premature expiry.
Storing large amounts of stock with little understanding of their demand may only lead to high levels of wastage, especially of goods that have a short shelf life. Unsold items mean capital that is held up in stock – capital that could have been put to good use in other business operations. Even if stock clearance sales were conducted to free warehouses the loss of revenue could be quite significant.
Limitations in inventory management with conventional forecasting methods
In most traditional methods or solutions demand forecasts are based on the current trend and seasonality observed over a few years. However, there could be other variables that are kept out of the equation but are important to understand consumer behavior.
The limited visibility on the demand drivers can lead to erroneous estimations. Even the adjustments with subjective calls and intuitive methods fail when the operational range expands. Owing to these characters the forecasts are likely to become increasingly inaccurate.
In order to accurately forecast future demand, you may have to look beyond conventional parameters such as historical trends and seasonality. Some of the influencers that need to be considered could be;
price, discounts, availability, delivery timeline and charges, etc.
Impact of external variables such as weather, rainfall, holidays, competitor pricing, inflation, etc.
Marketing spend & promotions such as ads, price-off, bundled offers, cash-backs
On top of accuracy-related challenges, e-retailers struggle with forecasting due to the sheer volume of items in their product portfolio. Most of the conventional methods do not lend to forecasting 1000s of items at once.
What are the solutions?
In order to achieve a high level of forecasting accuracies, it is important to use some of the latest modeling techniques that can consider a multitude of factors that influence the demand. Some of these techniques are grounded in machine learning algorithms while the rest are advance statistical models. Either way, many techniques need to be explored before zero in on one as they all have their strengths and weaknesses.
It is also important to have a process that can accommodate developing 1000s of forecasts within a reasonable timespan. This is precisely why a process with a high level of human intervention may not be suitable in this context.
Last but not least, the forecasts need to be dynamic and should lend to periodic re-calibrations.
Forecasting is just the beginning!
Having a forecasting process with the above qualities is a good place to start. But it is merely the beginning, not the end. You will have to make sure the forecasting is operationalized by incorporating it into your ordering and procurement process. This is how you can create business impact by way of accurate and large scale forecasting